Re: [v6ops] Geoff/queue-closed: 56%

[Brian Candler <[e-mail]>]

On 17/04/2024 12:08, Gert Doering wrote:
> On Wed, Apr 17, 2024 at 11:53:35AM +0100, Brian Candler wrote:
>> Also, everyone should be clear that there is currently no such thing as
>> "migrate to IPv6". You either add IPv6 to your IPv4 stack, or you stay with
>> IPv4. You can't remove the IPv4 part without cutting yourself off from most
>> of the Internet.
> This is not a message the IETF should be spreading, because then the logical
> question follows "why bother with IPv6 at all, if I have to keep IP4 around".

That is ultimately why this so-called "migration" hasn't happened - and 
wishful thinking won't change that, nor keeping quiet about reality. If 
there's any message the IETF should be spreading, it's that they live on 
the same planet as the rest of us.

Dual-stack would be excellent as a migration tool: add v6 this month, 
remove v4 next month, job done.  But that's not what's being asked of 
users, and dual-stack as a "forever" proposition is absolutely terrible. 
Nobody should be surprised that it has been rejected for the last 25 
years, and will continue to be rejected. Dabbling at the edges does not 
change this fundamental fact - such as saying "oh look, we measure lower 
RTT for IPv6 than IPv4, now you have a good reason to enable IPv6!". It 
isn't, and they won't.

We all wish there were more addresses, and this is most keenly felt at 
the access ISP side of things. In the UK, many new fibre operators are 
starting to put people behind CGNAT, which nobody wants.

As I see it, there are two general ways to approach this.

(1) You can try to get *all* the content providers in the world to 
enable IPv6.  This is the approach which has been tried and failed over 
 >25 years.

IMO, this was always doomed to fail. For as long as any significant 
banks, shops, restaurants or government services are not reachable via 
IPv6, then an IPv6-only connection is useless. And it's no good if 50% 
are, or 80% are, or even 99% are.  As long as that 1% is still there, an 
IPv6-only connection does not let you reach the Internet, and users will 
complain that the Internet is broken.  And as said above: "why bother 
with IPv6 at all, if I have to keep IP4 around?"

If there is no incremental benefit to users, neither is there 
incremental benefit to content providers. From where I'm looking, many 
large organizations aren't making their content available on IPv6 (e.g. 
www.bbc.co.uk - who were technical leaders in broadcasting, back in the 
20th century); neither are smaller ones, including technical 
publications like www.theregister.co.uk. Just two random examples.

Only they know the real reasons why they're not, but I believe it boils 
down to a negative business case: cost and time, with on-going running 
costs (dual-stack and happy-eyeballs makes network problems harder to 
diagnose), plus no return on investment, plus risks associated with user 
tracking, advertising and monetization, all set against higher 
priorities which *do* have a return on investment.

You can't tell people that their business case is wrong, or try to make 
up business cases for them. It's their money and their time, and they 
know themselves how best to allocate it. It's patronising to suggest 
otherwise, and they will rightly ignore you.

(Aside: it's also worth mentioning that there's no scarcity of IPv4 
addresses at the content provider and CDN side. They've been sharing IP 
addresses been sites for years.)

So that's failed. What's the alternative?

(2) You can make it practical to deploy IPv6-only networks at the access 
edge - the places where people sit behind NAT44 today - and still reach 
IPv4-only content. In that case, single-stack IPv6 replaces single-stack 
IPv4, and NAT64 replaces NAT44, and *that* starts to become a 
proposition you can sell: you are choosing between technology A or 
technology B, not technology A or technology A+B.

This is an approach the IETF could spread, with the support of big orgs 
like Google who are doing it in their own networks.

However, building NAT64 needs clue on the side of the user (if they are 
building their own NAT64), or on the side of the ISP (if they are 
running a large scale NAT64 on behalf of their users), and it has 
similar operational overhead to CGN/NAT44, and needs the same external 
public IPv4 resources, so it is still an uphill struggle.  It's a clear 
win over dual-stack, but much less so over IPv4 single-stack.

With my wishing hat on: I would like to see one or more big companies 
step up to the mark and build a giant, *public* NAT64 or reverse proxy 
so that the whole IPv4 Internet is accessible to anyone with an 
IPv6-only address. The big CDNs like Google, Cloudflare, Akamai etc 
would be well placed to do that.  At that point, IPv6-only access ISPs 
and enterprise networks spring into life.  I would expect usage to ramp 
up as challenger ISPs start to make use of this service; at that point, 
websites will see more and more of their traffic coming from the CDN 
source v4 addresses, and will realise they can get *better* tracking of 
their users by enabling native IPv6, at which point usage will decline. 
Incremental benefits to all. I would love this to happen, but I'm not 
holding my breath.

Regards,

Brian.